EVIDENCE OS / RESEARCH IN PROGRESS
Medical evidence, moving forward.
FEATURED · MEDICAL EVIDENCE
When agents help people choose, what makes your product or service credible?
Explore the service and deliverablesContinuously updated evidence from global clinical research
Clinical AGI decision-making infrastructure
What would you like to resolve?
Choose your context. See deliverables and fees.
Patients & families
Which treatment evidence fits this person?
3 scopes · Deliverables & examplesCliniciansNeed case appraisal or results from a blocked research stage?
2 scopes · Deliverables & examplesResearch teamsIs this topic worth pursuing, and how do we finish?
3 scopes · Deliverables & examplesHospitals & departmentsHow can our department use a shared evidence base?
2 scopes · Deliverables & examplesHealth brands, life sciences & CROsHow well are our product and health claims supported?
3 scopes · Deliverables & examplesLegal professionalsWhat supports the medical claims in this dispute?
2 scopes · Deliverables & examplesPayers, policy & guideline groupsHow do we compare effects, costs and implementation?
2 scopes · Deliverables & examplesEOS is the evidence system. Services are scoped by the work: General evidenceProfessional analysis & researchCustom engagement
EOS CLINICAL AGI RESEARCH INSTITUTE
Explore the Clinical AGI InstituteInstitute · Research
Research output
An auditable evidence system for LLM-assisted systematic reviews: development and internal evaluation
manuscript; development and internal evaluation
Explore the studyResearch output
Large language models in medical decision making: from clinical answers to living evidence
manuscript, not peer reviewed
Explore the studyResearch output
Large language models for patient-centred evidence generation and use
Author manuscript, not peer reviewed; proposed framework
Explore the studyIn progress
Musculoskeletal–metabolic comorbidity
A traditional systematic review and meta-analysis assisted by LLMs is in the review stage.
Progress & next stepsResearch programme and evolving plans11 directions
THE RESEARCH PROGRAMME
Research directions and next steps
Each result becomes a foundation for the next study. Explore progress, next steps and research basis.
Research reportedIn progressStudy designedPlanned
Research reported
Patient-centred evidence
How should evidence be organised around patient goals and comorbidity?
A patient-centred framework manuscript organises evidence generation and use around comorbidity, patient goals and concrete questions.
Next steps & research basis
Next
Translate the framework into measurable tasks and evaluation criteria.
Research reported
Research landscape of medical LLMs
Which steps from clinical answers to living evidence have been evaluated?
A published review manuscript examines evidence across medical dialogue, retrieval, synthesis, updating and individual decisions.
Next steps & research basis
Next
Refine the programme and its evaluation requirements through continuing literature tracking.
Builds onPatient-centred evidence
Research reported
Social connection and surgical outcomes · Evidence compilation
Can large literature collections become reviewable, reusable evidence?
The social-connection review contains 20,399 deduplicated records and 445 studies; 72 enter at least one quantitative synthesis.
Next steps & research basis
Next
Assess reuse in comorbidity research and retain evidence that cannot be pooled.
Builds onPatient-centred evidence
Research reported
Reliability, governance and correction
Can each conclusion be traced to its sources, revisions and responsibilities?
An auditable system was developed and internally evaluated in one review: 50 root-cause events, 46 resolved and 4 retained limitations.
Next steps & research basis
Next
Run independent comparative evaluations of errors, costs and reviewer workload.
In progress
Musculoskeletal–metabolic comorbidity
Which comorbidity evidence can inform clinical discussions?
A traditional systematic review and meta-analysis assisted by LLMs is in the review stage.
Next steps & research basis
Next
Complete conventional synthesis, then use the case to evaluate mixed-information synthesis and individual applicability.
Study designed
Mixed-information synthesis
How much information can be retained when effect estimates are incomplete or measures and time windows differ?
Method boundaries and an evaluation design compare conventional and mixed-information synthesis for coverage and credibility.
Next steps & research basis
Next
Test additional information, statistical coherence and clinical value in the comorbidity review.
Study designed
Evidence to individual decisions
How should synthesised evidence fit decisions by patients, clinicians and institutions?
A decision-material evaluation design is established; comorbidity evidence will support further evaluation of individual applicability.
Next steps & research basis
Next
Evaluate applicability, traceability, justified abstention and decision quality; confidence alone is insufficient.
Planned
Living evidence and updates
Which conclusions should change when new studies or corrections appear?
The programme includes update triggers, versioning and correction propagation.
Next steps & research basis
Next
Define change comparisons and review triggers; evaluate delay and maintenance costs.
Planned
Evidence gaps to new studies
How can unanswered clinical questions become the next study?
The plan connects gap detection, study design and research collaboration.
Next steps & research basis
Next
Select high-value gaps from existing reviews and assess topic selection and gap closure.
Planned
Independent comparison and external validation
Can the full system improve workflow, decision quality and patient outcomes?
The plan compares human, human–AI and AI workflows in independent domains, then progresses toward hospital evaluation.
Next steps & research basis
Next
Progress through benchmarks, prospective silent testing, workflow pilots and outcome evaluation.
Planned
Complex clinical decisions and foundational AI
Can comorbidity, uncertainty and value trade-offs advance decision models?
A future foundational research programme builds on empirical synthesis and decision-translation studies.
Next steps & research basis
Next
Complete comorbidity and method evaluations before defining falsifiable model-research questions.
FRONTIERS · DAILY BRIEF
Read daily insightsFrontiers
Understand advances in clinical AI and medical evidence, and what they mean for health choices and clinical practice.
Daily selection · compiled
AI-extracted health data should remain traceable to the source
The study compared manual entry with AI extraction of breast-cancer trial data from EHRs. Each extracted variable retained a source-text location for reviewer verification.
What it means for health and care
For clinicians and research clients, source-linked results make misreadings easier to detect, disagreements easier to resolve and corrections easier to record—an important safety condition beyond efficiency.
Study limits and provenance
This was a retrospective comparison in one hospital and one cancer type. It does not establish reliable performance across hospitals, lower costs or better patient outcomes.
Artificial intelligence for clinical data extraction from EHRs in breast cancer trials: the MIRROR study
Medical AI can be steered even when it cites evidence
The study shows that adding a few crafted documents to a retrieval corpus can change how medical AI frames the same facts and steer answers toward a product or viewpoint.
What it means for health and care
Patients and clinicians need more than the presence of citations: who produced the sources, whether independent sources agree, whether commercial framing is present and whether conclusions change when sources are replaced.
Study limits and provenance
This is a laboratory benchmark, not evidence that the same attack has occurred in clinical practice. Real-world frequency, patient harm and the best defence remain unproven.
FramePoison: attacks on medical RAG that target framing, not just facts
Very little published clinical evidence supports AI used during surgery
Researchers screened 3,020 records and found only five studies of real intraoperative decision support. Only one had completed results, based on five patients.
What it means for health and care
Patients, surgeons and hospitals should not treat technical demonstrations, registered trials or clinician final authority as proof of safety or benefit. Auditability, accountability and outcome evaluation remain necessary.
Study limits and provenance
The review was limited to intraoperative AI, included few heterogeneous studies, and proposed an externally unvalidated governance scorecard.
Ethical considerations for intraoperative implementation of artificial intelligence clinical decision support systems: a scoping review
METAQUANTCompare the evidence behind research investment.Research assets, ROI and trends
EVIDENCE PYRAMID
Evidence has levels. Decisions have context.
From global research to the individual patient: inspect how the evidence was produced and whether it answers the question.
A simplified guide for treatment-effect questions. Bias, consistency, precision and relevance determine confidence alongside design.
- Systematic reviews & meta-analyses
- Randomized controlled trials
- Observational studiesCohorts · registries · case–control
- Case reports & case series
- Expert opinion & mechanistic reasoning
Patient goalsCircumstancesIndividual relevance
Why choose Evidence OS?
CREDIBILITY
Clinical and research expertise
Clinical and research expertise combines original sources, professional review and real studies to synthesize evidence through systematic reviews and meta-analysis.
Chuan Yin · Doctorate in Clinical Medicine, Peking University; Visiting Scholar, Harvard Medical School; Founder and Chief Scientist, Evidence OS.CREDIBILITY · RELIABILITY
Advanced mathematical and statistical methods
Develop and apply advanced mathematical and statistical models for mixed-information evidence synthesis, comorbidity research and personalized analysis. Explain assumptions, applicability and uncertainty behind method choices.
RELIABILITY
Traceable findings and processes
Provide sources, data, analyses and versions within the agreed scope. Review, corrections and updates retain an evidence trail showing how findings were formed.
RELIABILITY
Clear delivery and update commitments
Agree scope, fees, responsibilities, milestones and acceptance before engagement. Track new studies as agreed and confirm additional work and fees in advance.
INTIMACY · UNDERSTANDING
Understand concerns and respect boundaries
Listen to your goals, concerns, budget and constraints, and confirm the question together. Explain evidence in context and agree how information will be used and handled.
CLIENT INTERESTS
Assess value before committing
Present supporting, opposing and insufficient evidence honestly. Start research commissions with an assessment, then choose to proceed, narrow or pause; credit reusable work as agreed.
- Describe your question
Describe the purpose, inputs and timing. Scoping requests are free.
- Agree scope & quote
Define deliverables, professional review, fees and acceptance.
- Work & deliver in stages
Begin after contracting; agree update needs separately.
GLOBAL RESEARCH · CLINICAL CONTEXT
Multiple evidence routes. One clinical question.
Journals, evidence syntheses, registries and clinical expertise complement one another.
1 / 6
NEJM
Clinical researchPeer-reviewed studies, reviews and clinical cases. Inspect the design, population, findings and relevance of each article.
Visit original sourceJAMA Network
General and specialty medicineClinical research, commentary and specialty journals provide traceable literature across medical questions.
Visit original sourceCochrane
Evidence synthesisSystematic reviews combine studies around focused questions, including benefits, harms and uncertainty.
Visit original sourceAccess to public material and full text depends on permission; expert contributions and project inputs require specific authorisation. Source names and marks do not imply institutional partnerships.
Explore sources and methods